France Passes Fast-Fashion Law Targeting Shein and Temu
France has become the first major European country to pass a law regulating ultra-fast fashion, with the Senate giving final approval on 29 June after the National Assembly cleared the text the week before. The bill, roughly two and a half years in the making, now awaits President Macron's signature. It does not ban ultra-fast fashion; instead it introduces an escalating per-item environmental fee - starting between €0.25 and €6 and rising toward €10 by 2030 - alongside a ban on advertising for ultra-fast-fashion brands and on influencers promoting them.
The law creates a formal legal definition of "ultra-fast fashion" based on how fast a company releases new products and how broad its catalogue is, plus a repairability test that penalises garments cheaper to replace than to mend. Crucially, that definition is drawn to capture the Asian platforms - Shein, Temu and AliExpress are named targets - while European players such as Zara, H&M and Kiabi fall outside it, prompting Green Party lawmakers and the Stop Fast Fashion coalition to attack the final text as watered down. The bill's sponsor, MP Anne-Cécile Violland, was candid about the intent: "We're coming down very hard on Shein, and that's the first step."
FRANCE 24
Why it matters: What France has legally defined here is not cheap clothing but a technology model - Shein and Temu run algorithmic, demand-tested supply chains that push huge SKU volumes through small-batch, on-demand production, and the "speed of release and breadth of assortment" test is drawn around precisely that, which is why Zara and H&M sit outside it despite comparable scale.
The compliance response will be a data problem before it is a sustainability one: per-item eco-modulation fees and the repairability metric have to be calculated at SKU level, pushing exactly the product-level traceability the EU's Digital Product Passport is already set to mandate. The near-term squeeze on ultra-low-cost retail is arriving through cost and data rather than the morality framing the debate leans on - France's €2 parcel fee, the EU's planned €3 import charge this autumn, and now an environmental levy that only functions if products carry structured data.
For brands, the message is that algorithmic fast fashion now has a statutory definition in a major market, and meeting it will be an infrastructure question first.
New Look adopts Fermat to cut sampling from its design process
New Look has partnered with the generative AI platform Fermat to equip its buying and design teams with tools that turn sketches into photorealistic renders before anything is physically made. Fermat, a Barcelona-based startup, lets designers generate virtual product visuals, test multiple iterations and explore prints, colourways and styling options, with the stated aim of reducing reliance on traditional sampling.
The retailer frames the tie-up as part of a wider digital transformation that also draws on data from its Club New Look loyalty programme to inform what gets designed. It is the second significant UK fashion name to sign with Fermat this year, after ASOS embedded the tool across its design function in February. Fermat markets time and cost savings of 75 to 80% on core design tasks, though that is the vendor's own figure, drawn from earlier deployments rather than anything New Look has reported.
Scott Thompson
Why it matters: The announcement leans on the language of creativity, but the commercial logic is cost. New Look closed 20 stores in the year to March and is reportedly weighing a sale, and for a retailer in that position the draw of Fermat is the sampling spend it removes, not the imagination it unlocks. The ASOS deal two months earlier makes Fermat an emerging default for sketch-to-render work among UK mid-market fashion, which is noteworthy.
Virtual Try-On's Body-Image Problem
A consumer-magazine analysis has put a sharper question to virtual try-on: what if the body it shows you back is not quite your own? The piece argues that AI try-on tools embedded in fashion retail are returning subtly altered versions of users' uploaded photographs - waists narrowed, limbs smoothed, hip lines redrawn - and that this is a feature of how the technology works rather than a fault.
The mechanism is the substantive part. Contemporary try-on increasingly relies on generative models that do not simply drape a garment over a photo but reconstruct body regions the system cannot read cleanly, such as obscured limbs or overlapping fabric, and that reconstruction draws on training data skewed toward thin, young, standardised shapes. Because the edit begins from the shopper's real image and nudges it only slightly, the result is plausible in a way a superimposed model body is not.
The analysis situates this against research linking repeated exposure to idealised, manipulated imagery to lower body satisfaction in young women, and cites fashion historian Emma McClendon on what is new here: the manipulation is applied to your own image, on your own device, at the point of purchase, rather than to pictures of other people. Notably, the piece names no specific tool or platform, and points to little in the way of regulation beyond a comparison to TikTok's ban on beauty filters for under-18s.
Jade McGee
Why it matters: The sourcing is a little thin - no named products, and body-image research drawn from social feeds rather than try-on itself - so the claim that retail tools are systematically slimming shoppers should be read as argument, not established finding. But the mechanism underneath it is real and auditable, and it is where product teams should pay attention: as try-on shifts from garment-warping to diffusion-based generation, these systems reconstruct the body, and a reconstruction inherits whatever the training set treats as a default shape. That is a design decision, not a glitch.
The commercial exposure is the sharp end of it - a plus-size shopper shown a slimmed version of herself is both a trust problem and a self-defeating one, because if the rendered body is not the real body, the fit claim is worthless (at least, unreliable) and the return the tool was meant to prevent still arrives at the door. The same generative capability that makes try-on cheap to scale is what introduces the body-fidelity risk, and vendors that can demonstrate they render the actual customer, rather than an idealised proxy, will have something worth selling.
